CoCo Tool

CoCo Tool is a tool for data extraction for AI training.

• -connectionString - connection string to DB where input data are stored.
• -layer - input data table name.
• -classColumn - column name where saved class names. For the extraction use only unique names.
• -annotationColumn - column name where bbox with annotation is stored.
• -mappingFile - metadata file where unique class names are listed.
• -metadataFormat - format of output metadata file. If you skip this attribute metadata will be stored in default json format.
• -imageSize - size for output image size in px.
• imageOverlap
  • -imageOverlapX - overlap of image height in px.
  • -imageOverlapY - overlap of image width in px
• AOI - here we can describe AOI (area of interest). That means we can ignore image tiles by giving application start and end positions in px from where to start to extract tiles.
  • -aoiStartX
  • -aoiStopX
  • -aoiStartY
  • -aoiStopY
• annotationsize - if annotation smaller then given size in px it will be extended.
  • -minAnnotationWidth
  • -minAnnotationHeight
• -outfolder - location for results

• -egidbProject
• ?
  • -negativeStarX
  • -negativeStopX
  • -negativeStarY
  • -negativeStopY

* -egidbMapping

Getting started with CoCo is very easy. You can call the CoCo Tool by using the command-line. Open the command-line window in the folder where the CoCo Tool files are stored on your machine or navigate to the location by using CLI. Then use the following command.

coco -cmd annotationtococo -connectionString "database=yourdbname;User Id=postgres;password=admin;host=localhost;port=5432" -layer "public.tablenamewithdata" -classColumn "element" -annotationColumn "bbox" -mappingFile "C:\path\to\mappingfile.txt" -metadataFormat "csv" -imageSize "512" -outfolder "C:\path\to\output" -imageOverlapX "384" -imageOverlapY "384" -minAnnotationWidth "32" -minAnnotationHeight "32" -aoiStartX 0 -aoiStopX 8192 -aoiStartY 0 -aoiStopY 4096

Let's walk true command:

• coco - is the name of the program.
• -cmd - flag after which you place arguments
• annotationtococo - name of task which extracts data for AI training